docs(agents): work directly and proportionately - #88
Conversation
…tive Ten rules that bound how an agent spends a turn: read what the change needs, implement the scope asked for, validate once, stop at the acceptance criteria. Written after a workstream that ran four sessions on spec amendment. CLAUDE.md pairs it with the anti-creep line on model reassessment: the licence to refactor past work is not a licence to grow the conversation. Claude-Session: https://claude.ai/code/session_01JZFxedvD7EtDfLFbNKodZt
|
Warning Review limit reached
Next review available in: 19 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 947f391347
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| UBIQUITOUS*LANGUAGE.md · dead end → the subsystem's C-section · how to \_use* | ||
| clave → README.md · how to _work on_ clave → CONTRIBUTING.md. Not here. This |
There was a problem hiding this comment.
This changes the routing instruction from UBIQUITOUS_LANGUAGE.md to the nonexistent UBIQUITOUS*LANGUAGE.md and leaves emphasis spanning through \_use*, so rendered instructions lose the underscore and italicize unrelated text. Restore the original filename and balanced *use* markup so agents can follow the documented knowledge-routing path.
AGENTS.md reference: AGENTS.md:L24-L26
Useful? React with 👍 / 👎.
| Work directly and proportionately. | ||
|
|
||
| - Read only the files necessary to complete the requested change. |
There was a problem hiding this comment.
The new rules appear only in the explicitly Claude-specific file, despite this project supporting other CLI-based agents and the commit stating that the directive was landed in both agent-facing index files. Any Codex or future agent that reads only AGENTS.md therefore receives none of the requested scope controls; move or duplicate the shared block there.
AGENTS.md reference: AGENTS.md:L8-L10
Useful? React with 👍 / 👎.
| - Make ordinary engineering choices without asking. | ||
| - Ask a question only if different interpretations would materially change the implementation. | ||
| - Do not perform separate review, re-verification, or subagent validation passes unless requested or a test fails. | ||
| - Run the smallest relevant validation suite once after changes. |
There was a problem hiding this comment.
For Claude-authored PRs, directing the agent to run only the smallest relevant suite conflicts with the repository requirement that every PR show all four gates green. Because CLAUDE.md is the more agent-specific instruction, this can cause Claude to skip workspace tests, the WASM build, or clippy even when those required checks would expose a regression; explicitly retain the full gate suite for PR changes.
AGENTS.md reference: AGENTS.md:L54-L61
Useful? React with 👍 / 👎.
Docs only, no overlap with this branch's files. Claude-Session: https://claude.ai/code/session_01JZFxedvD7EtDfLFbNKodZt * origin/main: docs(agents): work directly and proportionately — Ollie's scope directive (#88)
1 participant
Ollie's scope directive, landed into both agent-facing index files.
Ten rules bounding how an agent spends a turn — read what the change needs,
implement the scope asked for, validate once, stop at the acceptance criteria.
CLAUDE.mdpairs them with an anti-creep line on the model-reassessmentparagraph: the licence to refactor past work is not a licence to grow the
conversation.
Docs only. No code, no gates affected.
Note: the ten-rule block is currently duplicated across
AGENTS.mdandCLAUDE.md(which@-includes the former). Deliberate or worth collapsing toone copy — Ollie's call, tracked here rather than silently changed.
https://claude.ai/code/session_01JZFxedvD7EtDfLFbNKodZt